Parameter Selection for Digital Satellite Model Updating with Knowledge Graph

Digital twin technology has played an indispensable role in the recent satellite life cycle, specifically in design, testing, and operation monitoring. A digital twin is created according to a physical satellite and can mirror its performance and parameters. The physical satellite is occasionally affected by dynamic disturbances when in orbit, causing inconsistencies with its digital twin, thus requiring the digital twin model to be continuously updated through model parameter adjustments. Nonetheless, the digital twin model involves numerous parameters with coupled relationships; adjusting all simultaneously might be impractical as it requires high computational and time. Since only some of the model parameters have a significant impact on the digital model's outputs, we propose a parameter selection method that utilizes knowledge graph technology for digital satellite model updating. Firstly, the satellite state variables and their related telemetry parameters of satellite subsystems in the digital satellite model are defined. Secondly, a knowledge graph consisting of the correlation and sensitivity relationships of the state variables and telemetry parameters is built. Thirdly, a clustering algorithm and graph reasoning algorithms are employed on this knowledge graph to select and group the state variables and telemetry parameters required in model updating. The results show that by updating the state variables and telemetry parameters selected from the knowledge graph, not only can the state variables converge to optimal values in fewer iterations, but a more accurate digital satellite model can also be obtained.